Mistral's New Fine-Tuning Tools

Today:

• Mistral's New Fine-Tuning Tools

• Stability AI Introduced Stable Audio Open

• Cohere Raises $450M in Funding

• Musk Redirects AI Chips to X

• Asana Unveils AI for Teams

• Cisco Invests $1B in AI

• AI Predicts New Antibiotics

• Nvidia Tops Apple at $3.01 Trillion

• OpenAI Secures AI Research Infrastructure

• ChatGPT Recovers from Major Outage

• Apple, OpenAI Join Forces in AI

• Humane Issues Urgent Charging Case Alert

Mistral launches fine-tuning tools to make customizing its models easier and faster

Mistral, an open-source AI model provider, has launched new tools for fine-tuning its models, making customization faster and cheaper. These tools are available on their platform, La Plateforme, and allow for efficient fine-tuning that lowers costs. 

Mistral also offers a lightweight codebase, mistral-finetune, for fine-tuning on users' own infrastructure, and serverless fine-tuning services. Their models, including the 7.3B parameter Mistral 7B, can be customized through Mistral's API. The company has quickly risen, securing significant funding and partnerships, and continues to innovate in AI model development and customization.

VENTUREBEAT

Introducing Stable Audio Open - An Open Source Model for Audio Samples and Sound Design

Stable Audio Open, a new open-source model from Stability AI, generates high-quality audio samples up to 47 seconds from text prompts. Ideal for drum beats, instrument riffs, ambient sounds, and sound effects, it allows users to fine-tune the model with custom audio data. This differs from the commercial Stable Audio product, which creates full tracks up to three minutes with advanced features. 

Trained on Freesound and Free Music Archive data, Stable Audio Open emphasizes responsible development and creator rights. Available on Hugging Face, it invites musicians and sound designers to explore and provide feedback.

STABILITY AI

Nvidia and Salesforce double down on AI startup Cohere in $450 million round, source says

Canadian AI startup Cohere raised $450 million in funding from Nvidia, Salesforce Ventures, Cisco, and PSP Investments. This funding round values Cohere at $5 billion, a significant increase from its previous $2.2 billion valuation. 

Cohere specializes in generative AI models for enterprises with a focus on data privacy, generating $35 million in annualized revenue by March. The company, founded in 2019, competes with OpenAI, Anthropic, and Mistral. It benefits from the Canadian government's investment in AI research. Cohere aims to secure between $500 million and $1 billion in total funding.

REUTERS

Elon Musk ordered Nvidia to ship thousands of AI chips reserved for Tesla to X and xAI

Elon Musk instructed Nvidia to prioritize AI chip shipments to his companies X (formerly Twitter) and xAI over Tesla, delaying Tesla’s receipt of $500 million worth of processors by months. This decision impacts Tesla's plans for AI and robotics development. 

Nvidia emails reveal Musk's actions, leading to conflicts with Tesla shareholders concerned about his divided attention among multiple ventures. Tesla shares dropped 1% following the news. Despite Musk's claims of significant AI spending and infrastructure plans, Nvidia’s internal communications highlight discrepancies in Tesla’s procurement and operational readiness.

CNBC

Asana Unveils AI Teammates to Tackle Complex Workflows and Elevate Teamwork

Asana has introduced AI teammates to enhance team workflows and productivity. These AI collaborators, built on Asana’s Work Graph® data model, offer customizable support by advising on priorities, managing tasks, and improving workflows. They provide transparency and control, working alongside humans to ensure efficient task management. AI teammates are currently in beta, helping with tasks such as creative production and marketing campaigns. 

The launch also includes a new AI chat interface for intelligent assistance. Asana's AI aims to boost productivity and innovation, with customers like Palo Alto Networks and Anthropic envisioning significant benefits from integrating AI into their workflows.

ASANA

Cisco Launches $1B Global AI Investment Fund

Cisco has launched a $1 billion global AI investment fund to foster the development of secure and reliable AI solutions. The fund aims to support startups like Cohere, Mistral AI, and Scale AI, helping to build a robust AI ecosystem. Cisco has already committed $200 million to this initiative. 

The company emphasizes the importance of integrating AI to enhance products and services, aligning with their strategy to connect and protect the AI era. Cisco plans to collaborate on AI products, ensuring they meet customers' needs while maintaining security and reliability. This initiative underscores Cisco's dedication to AI innovation and strategic growth.

CISCO

AI used to predict potential new antibiotics in groundbreaking study

A groundbreaking study has used AI to predict new antibiotics, marking a major advance in antibiotic resistance research. Scientists at the University of Pennsylvania, led by César de la Fuente, developed an algorithm to analyze the global microbiome, discovering nearly one million new molecules with potential antimicrobial properties. This AI-driven method significantly accelerates the discovery process compared to traditional techniques. 

The research, published in Cell, identified nine potential antibiotics from 6,680 compounds, addressing the urgent public health issue of antimicrobial resistance. The study's data and code are publicly available to further scientific progress and benefit humanity.

THE GUARDIAN

Nvidia is now more valuable than Apple at $3.01 trillion

Nvidia has surpassed Apple to become the second most valuable company globally, with a market capitalization of $3.01 trillion. This milestone comes as Nvidia's dominance in the AI sector, particularly with its H100 chip, drives its market value higher. The company reported $14 billion in profit from AI chip sales in May and now trails only Microsoft, which has a market cap of $3.15 trillion. 

Nvidia's stock price has surged to over $1,220 per share, and the company plans a stock split on June 7. Nvidia aims to release a new AI chip annually, starting with the Blackwell B200 GPU.

THE VERGE

Securing Research Infrastructure for Advanced AI

OpenAI has revealed its secure architecture for training advanced AI models on their research supercomputers, emphasizing security and efficiency. Built on Azure and managed with Kubernetes, the system prioritizes protecting sensitive data, particularly unreleased model weights. Key features include robust identity management with Azure Entra ID, strict Kubernetes role-based access control, and a sophisticated identity and access management service called AccessManager. 

The infrastructure employs a defense-in-depth approach with modern VPNs, egress controls, and continuous auditing and testing. OpenAI also uses internal and external red teams to simulate threats and enhance security measures continually.

OPENAI

ChatGPT is back online after an outage

OpenAI's ChatGPT experienced two major outages on Tuesday, affecting both web and mobile users. The first outage occurred early in the morning, followed by a second one around 10:30 AM ET, which lasted until 1:17 PM ET. 

During this period, users reported issues accessing ChatGPT, as well as AI tools like Google Gemini, Anthropic’s Claude, and Perplexity AI. These issues were later resolved. Previous outages, such as one in November caused by a DDoS attack, also impacted OpenAI’s API services. The recent outages did not affect OpenAI's API.

THE VERGE

Apple Made Once-Unlikely Deal With Sam Altman to Catch Up in AI

Apple is partnering with OpenAI to advance its generative artificial intelligence capabilities. This collaboration will be officially announced at Apple's upcoming Worldwide Developers Conference (WWDC) in 2024. The alliance aims to help Apple catch up in the AI field, leveraging OpenAI's expertise. 

This partnership marks a significant step in Apple's AI strategy, reflecting its ongoing efforts to enhance AI-driven features across its products, including AI-based enhancements for Siri and AI-focused M4 chips. Sam Altman, CEO of OpenAI, previously engaged with Apple in 2008, and now returns to support Apple's AI ambitions.

BLOOMBERG

Humane warns AI Pin owners to ‘immediately’ stop using its charging case

Humane has advised owners of its AI Pin to stop using the charging case immediately due to a fire safety risk linked to a third-party battery cell. The company identified the issue after receiving a report about a charging problem and has since disqualified the battery vendor. The AI Pin itself, its Battery Booster, and the charging pad are not affected. 

As a compensation, Humane is offering two free months of its subscription service. The company is searching for a new battery supplier and will provide more information once the investigation is complete.

THE VERGE